NeurIPS 2020

Reward-rational (implicit) choice: A unifying formalism for reward learning

Meta Review

There were initial concerns from the reviewers about the contribution of this framework and whether it really changes what our fundamental understanding of the problem or practical capabilities, or if is more of a literature review with an interesting but not particularly impactful framework / unifying theory. After the author response and discussion, the reviewers came to the conclusion that this framing does, in fact, increase our understanding into how to compare many existing algorithms / feedback types, and that the meta choice model is a sufficiently novel and useful insight.